Morphometric Indices for Sexing Adult Royal Eudyptes Schlegeli and Rockhopper E

1996 Hull: Morphometric indices for sexing penguins 23

MORPHOMETRIC INDICES FOR SEXING ADULT ROYAL EUDYPTES SCHLEGELI AND ROCKHOPPER E. CHRYSOCOME PENGUINS AT MACQUARIE ISLAND

CINDY L. HULL

Department of Zoology, University of Tasmania, GPO Box 252C, Hobart, Tasmania 7001, Australia ([email protected])

Received 8 November 1995, accepted 4 March 1996

SUMMARY

HULL, C.L. 1996. Morphometric indices for sexing adult Royal Eudyptes schlegeli and Rockhopper E. chrysocome Penguins at Macquarie Island. Marine Ornithology 24: 23–27.

Four measurements were taken from a sample of known-sex Royal Eudyptes schlegeli and Rockhopper E. chrysocome Penguins at Macquarie Island. Significant differences were found between sexes in all measurements. Bill depth and bill length were the most reliable measures for assigning sex, and when these were applied to a discriminant function analysis, accurately assessed 97% males, 97.2% females in Royal Penguins (97.1% overall), and 93.3% males, 93.0% females in Rockhopper Penguins (93.2% overall). Cross-validation using jackknife analysis accurately assigned the sex of 97% of males, 97% of females in Royal Penguins (97.1% overall), and 93% males, and 93% females in Rockhopper Penguins (93.2% overall), indicating the validity of using these measurements. The non-overlapping ranges (mm) were: in Royal Penguins bill depth, males 29.3–33.5, females 24.3–27.7; bill length, males 67.0–73.8, females 55.8–62.7; in Rockhopper Penguins bill depth, males 20.3–23.3, females 16.7–18.8; bill length: males 46.2–51.2, females 38.5–42.2. These ranges should be used to assign sex in the field. For penguins that fall outside these ranges bill depth and length should be applied to the derived discriminant formulae. Some significant morphometric differences were found between Royal Penguins in this and a previous study on Macquarie Island, indicating the difficulty of com- paring studies involving different workers. Methods for overcoming these difficulties are discussed.

INTRODUCTION hopper Penguins, and to thus determine the most reliable means of sexing birds in the field. These data are then com- Studies of the ecology of avian species often require the iden- pared to previous studies of Royal and Macaroni E. chryso- tification of the sexes of individuals. In penguins there are no lophus Penguins. reliable plumage differences that can be used to distinguish the sex of individuals visually (Davis & Spiers 1990, Marchant & Higgins 1990). In order to avoid destructive or invasive tech- MATERIALS AND METHODS niques, the use of external morphometrics to sex animals reliably is of great value. Whereas penguins are dimorphic in The study was carried out at Macquarie Island (54°30'S, body mass, this measure is unreliable due to its variability 158°57'E) during the 1993/4 and 1994/5 breeding seasons. across animals within and between years (Warham 1975, Davis Royal Penguins were measured at Sandy Bay, and Rockhopper & Spiers 1990, Groscolas 1990, C.L. Hull unpubl. data). Penguins at Brothers Point, at the southern end of Sandy Bay. Measurements (to 0.1 mm) were made with Vernier calipers External morphometric indices have been used widely to assist of bill depth (at a point proximal to the tip of the triangular in the sexing of penguins (Scolaro 1987, Gales 1988, Kerry et inter-ramal feather patch), bill width (maximum width of the al. 1992, Amat et al. 1993, Agnew & Kerry 1995, Woehler culminicorn), bill length (length of exposed culmen) (as per 1995). However, there are no published data on the mor- Warham 1972, 1975) and head length (maximum length from phometrics of Rockhopper Penguins Eudyptes chrysocome the dorsal brain case to tip of the bill) of penguins of presumed filholi from Macquarie Island, and the few data for Royal Pen- sex as determined by breeding behaviour. Breeding behaviour guins E. schlegeli indicate a need for further statistical analy- included date of return to the island (males return at least one sis of morphometric characters (Woehler 1995). week earlier than females), incubation shift (females carry out the first incubation shift), and guard-stage foraging shifts Sexual dimorphism has been noted in all species of penguins, (undertaken by females) (Warham 1963, Smith 1970, Warham with males always larger than females (Livezey 1989). The 1971, Carrick 1972, Warham 1972, Marchant & Higgins degree of dimorphism, however, varies between groups of 1990). As there is no published, or observed, evidence of penguins. Using skin measurements Livezey (1989) found reverse-role behaviours in these species this technique was crested penguins Eudyptes spp. to be the most dimorphic of deemed to be reliable. Individuals were marked with perma- penguins, but they were only moderately dimorphic when nent metal flipper bands and were observed at least once per compared using skeletal measurements. week throughout the breeding season, enabling further confir- mation of the sex of an individual (Hull & Wilson 1996). The purpose of this paper is to provide data on morphometric During the 1993/4 season, 50 pairs of breeding Royal and 50 indices for identifying the sex of individual Royal and Rock- pairs of breeding Rockhopper Penguins were measured. Pairs 24 Hull: Morphometric indices for sexing penguins Marine Ornithology 24

TABLE 1 MORPHOMETRIC INDICES IN MM (MEAN AND STANDARD DEVIATION) OF ROYAL AND ROCKHOPPER PENGUINS FROM MACQUARIE ISLAND. ALL SIGNIFICANTLY DIFFERENT, P<0.05. Species Sex Bill depth Bill width Bill length Head length BSI * Royal Penguin male (67) 30.4 (1.57) 14.1 (1.23) 68.7 (2.85) 143.5 (4.72) 2954.3 (373.1) female (71) 26.8 (1.27) 13.3 (1.15) 61.1 (2.63) 133.8 (4.34) 2181.6 (246.8) t values 14.6 4.1 16.2 12.5 14.3 % difference between sexes 88.3 94.1 88.9 93.3 73.8 Rockhopper Penguin male (60) 21.0 (0.99) 10.8 (0.81) 46.4 (2.05) 115.6 (2.98) 1051.8 (114.6) female (57) 18.7 (0.85) 10.0 (0.85) 41.9 (2.13) 109.6 (3.63) 788.2 (116.4) t values 13.8 4.9 11.6 9.8 12.3 % difference between sexes 88.7 92.9 90.3 94.7 74.9

* Bill Shape Index

on nests were selected haphazardly from three transects in the character and (xm+xf) xf is the mean of the female character; Royal Penguin colony and two in the Rockhopper Penguins colony (see Hull & Wilson 1996). All birds were measured S = 1–p where p is the proportion of individuals that are while on the nest to minimise disturbance (Hull & Wilson misclassified by a single factor discriminant analysis; 1996). During the subsequent season, previously unbanded breeders on the transects were measured. Therefore, a total of BSA = prl where l is bill length and r is (=0.5 bill depth) 138 Royal (67 males, 71 females), and 117 Rockhopper (60 (the formula is the shape of a cone). males, 57 females) Penguins were measured.

Comparisons of morphometric data between the sexes were RESULTS made using Student t tests. Discriminant Function Analyses (DFA) were used to determine the accuracy of assigning pen- Sexing penguins by morphometric indices guins to a sex using these morphometric data, and to determine the most reliable measurements. A jackknife analysis was then All mean measurements in this study were significantly differ- used to cross-check the accuracy of the DFA (Tabachnick & ent between the sexes in both species (t-tests P<0.05), with Fidell 1989). From these results, discriminant formulae were males being larger (Table 1). The mean difference ranged from derived to assign a sex to individuals for future studies. In 73.8% (BSI) to 94.1% (bill width) in Royal Penguins, and addition, a Bill Shape Index (BSI) was calculated from the 74.9% (BSI) to 94.7% (head length) in Rockhopper Penguins multiplication of bill depth, bill width and bill length, and (Table 1). divided by 10 (see Warham 1975) for the purpose of comparisons with Woehler (1995). A Mean Dimorphism Index (MD), In Royal Penguins the canonical loadings from the DFA a Separation Index (S), and a Bill Surface Area (BSA) were determined that bill length was the most reliable predictor of also calculated for each of the characters for comparison to sex, and in Rockhopper Penguins, bill depth (Table 2). Lower other studies (see Agnew & Kerry 1995). These indices were loadings for bill width and head length in both species indi- defined as follows: cate a lesser contribution of these measurements to the accu- rate assignment of sex to a penguin. Using the four variables,

MD = 200 (xm–xf) (%) where xm is the mean of the male the discriminant function analysis accurately assessed 95.5%

TABLE 2 CANONICAL LOADINGS OF MORPHOMETRIC INDICES FOR ROYAL AND ROCKHOPPER PENGUINS FROM THE DISCRIMINANT FUNCTION ANALYSIS Variable Royal Penguins Rockhopper Penguins Bill depth 0.743 –0.883 Bill width 0.207 –0.318 Bill length 0.818 –0.749 Head length 0.633 –0.637

Group Classification function coefficients: Variable Males Females Males Females Bill depth 46.137 34.313 76.349 53.613 Bill width 125.164 121.261 73.915 75.703 Bill length 35.303 27.955 –20.634 –27.904 Head length 59.340 58.022 101.552 101.891 1996 Hull: Morphometric indices for sexing penguins 25 of males, and 97.2% of females in Royal Penguins (96.4% TABLE 3 overall), and 93.3% of males and 93.0% of females in Rock- hopper Penguins (93.2% overall). However, removing bill NON-OVERLAPPING RANGES FOR BILL DEPTH width and head length increased the accuracy of the DFA to AND LENGTH IN ROYAL AND 97.0% males, 97.2% females in Royal Penguins (97.1% over- ROCKHOPPER PENGUINS all), and 93.3 of males, 93.0% of females (93.2% overall). Species Sex Bill depth Bill length Cross-validation of these two variables using a jackknife analysis found that 97% of males and 97% of females in Royal Royal Penguins (97.1% overall), and 93% of males, and 93% of Penguins Males 29.3–33.5 (77.6%) 67.0–73.8 (73.1%) females in Rockhopper Penguins (93.2% overall) were accu- Females 24.3–27.7 (73.2%) 55.8–62.7 (77.5%) rately assigned a sex. The DFA produced the following formulae for the determination of sex in these penguins: Rockhopper Penguins Males 20.3–23.3 (75%) 46.2–51.2 (60%) Royal Penguins Females 16.7–18.8 (49%) 38.5–42.2 (63%)

D = –919.9 + (13.45 BD) + (8.24 BL) The figures in parentheses represent the proportion of animals that can be sexed using these ranges. Rockhopper Penguins

D = –739.3 + (21.97 BD) + (6.86 BL) MD and S values for all measurements are given in Table 5. Both the S values and the canonical loadings values given in Where BD = Bill Depth, BL = Bill Length, and D is the dis- Table 2 confirm that bill depth and bill length are the most criminant function. Using these formulae, individual penguins reliable measurements in both species. Because it was not pos- that fall above zero are male and those that fall below are sible to compare the MD and S values on BSI between this female. study and that of Agnew & Kerry (1995) and Woehler (1995), these were not calculated. Removing data that overlap between the sexes in each species, the non-overlapping ranges for bill depths and lengths for each of the sexes (mm) are given in Table 3. DISCUSSION

Inter-population comparisons Sexing penguins by morphometric indices

Royal Penguin measurements obtained in this study were com- The significant differences found in bill depth, bill width, bill pared with those of Woehler (1995). Bill depth of males, and fe- length and head length between the sexes of Royal and male and male bill length were found to be significantly differ- Rockhopper Penguins on Macquarie Island are of little sur- ent, with bill depth being less, and bill length being greater in prise. The dimorphic nature of penguins has been well docu- this study compared to Woehler’s (1995) (two-tailed t-test, mented (e.g. Warham 1975, Gales 1988, Livezey 1989, Davis P<0.05) (Table 4). BSI differed in opposite directions for males & Spiers 1990, Murie et al. 1991, Agnew & Kerry 1995), and females. Bill depth, bill width, bill length, and BSI were also although the extent of the difference in the Macquarie Island compared between Macaroni Penguins from Heard Island populations was previously unknown for Rockhopper Pen- (Woehler 1995) and Royal Penguins (this study). Significant dif- guins, and less well known for Royal Penguins. More impor- ferences were found in all measurements, with Royal Penguins tantly, it is now possible to determine the sex of individual being larger in both sexes (two-tailed t-test, P<0.05) (Table 4). penguins of these species without invasive techniques.

TABLE 4 MEAN (STANDARD DEVIATION) FOR MORPHOMETRIC INDICES OF ROYAL AND MACARONI PENGUINS AT MACQUARIE AND HEARD ISLANDS. (* P<0.05) Measurement Sex Royal Penguins Royal Penguins Macaroni Penguins 1 × 21 × 3 Macquarie Is. Macquarie Is. Heard Is. Significance Significance This study Woehler (1995) Woehler (1995) (% difference to 1) (% difference to 1) 12 3 Bill depth M 30.4 (1.57) 32.9 (1.87) 27.5 (0.82) * (92.51) * (110.62) F 26.8 (1.27) 27.6 (2.42) 24.0 (1.02) ns (97.24) * (111.69) Bill width M 14.1 (1.23) 14.7 (1.06) 12.9 (0.87) ns (95.99) * (109.37) F 13.3 (1.15) 13.1 (0.85) 11.1 (0.41) ns (101.84) * (119.30) Bill length M 68.7 (2.85) 64.8 (3.78) 61.4 (1.68) * (106.04) * (111.90) F 61.1 (2.63) 57.3 (3.34) 53.7 (2.07) * (106.69) * (113.68) BSI M 2954.3 (373.1) 3136.6 (391.2) 2166.2 (186.5) ns (94.19) * (136.38) F 2181.6 (246.8) 2078.4 (370.0) 1439.9 (95.7) ns (104.96) * (151.51) Sample size M 67 10 10 F71 10 10 26 Hull: Morphometric indices for sexing penguins Marine Ornithology 24

TABLE 5 MEAN DIMORPHISM INDEX (MD) AND SEPARATION INDEX (S) FOR ROYAL AND ROCKHOPPER PENGUINS Species Bill depth Bill width Bill length Head length BSA Royal Penguin 12.47 (91.30) 6.06 (60.14) 11.67 (92.75) 6.96 (85.50) 12.07 (96.38) Rockhopper Penguin 11.78 (92.31) 7.31 (67.52) 10.17 (88.03) 5.38 (86.33) 10.96 (93.16)

Discriminant function analyses indicated that the use of only pretation of precisely where on the bill measurements should two of the measurements is sufficient to assign accurately the be taken. This indicates that bill length, and to an extent bill sex to individuals of Royal and Rockhopper Penguins on depth, were more variable between workers than some other average 97.1% and 93.2% of the time, respectively. This rate measurements, with BSI being the most reliable measure. is higher than that of other species subjected to the same pro- Although BSI is a derived index, the differences between meas- cedure, such as the Adélie Penguin Pygoscelis adeliae, in urers of these variables deviates in different directions, result- which a 85% success rate was recorded, the latter being less ing in a masking of inconsistencies. The results also suggest dimorphic than the crested penguins in these measurements differences in all measurements between Royal (this study) and (Kerry et al. 1992). Macaroni Penguins (Woehler 1995). Due to the above difficulties these differences have to be viewed tentatively. There- The results from both the DFA and S (Single factor discrimi- fore, comparative studies with data derived by different work- nant analyses) values further confirm that the most reliable ers should be conducted with caution, particularly studies single morphometric measure for assessing sex was bill length describing inter-population variation within species. These in Royal Penguins, and bill depth in Rockhopper Penguins. findings confirm the work of Barrett et al. (1989) who suggest The S values (Table 5) indicate the percentage of individuals that all workers should measure the same birds to resolve dif- which did not overlap. Bill width was the least reliable meas- ferences in techniques, or a number of samples from various ure, with only 60.1% of measurements and 67.5% of measure- measurers be taken in all cases. ments not overlapping, and BSA the most reliable at 96.4% and 93.2% of measurements not overlapping in Royal and Rock- hopper Penguins, respectively. The non-overlapping ranges of ACKNOWLEDGEMENTS bill depth or bill length given above, or preferably BSI, can therefore, be used in the field to sex these species of penguin. This work was carried out with grants from the Antarctic Measurements from individuals that fall outside the ranges Scientific Advisory Committee and Sea World, Queensland, given, should be applied to the discriminant formulae pre- Australia. Thanks go to Jane Wilson and Mary-Anne Lea for sented above. The derived discriminant formulae cannot be assistance in the field, and to Kathy Haskard for statistical applied to chicks or juveniles, which in this and other penguin advice. Mark Hindell, Graham Robertson and Jan Murie com- species, have smaller bills than do adults (Warham 1972, Gales mented on a draft of the manuscript, for which I thank them. I 1988, Scolaro 1987, C.L. Hull pers. obs.). also thank Eric Woehler and an anonymous referee for their comments. The work was carried out under Macquarie Island Both Royal and Rockhopper Penguins from this study fall Nature Reserve Special Permits, numbers MI/34/94 and MI/ within the range of the mean dimorphism indices of the other 3/95. species of crested penguins given by Agnew & Kerry (1995). Rockhopper Penguins from this study were smaller in mean bill depth and mean bill length than in the other studies listed by REFERENCES Agnew & Kerry (1995). It has not been possible to compare data from this and other studies further statistically. However, using AGNEW, D.J. & KERRY, K.R. 1995. Sexual dimorphism in these morphometric indices (but not using other skeletal meas- penguins. In: Dann, P., Norman, I. & Reilly, P. (Eds). The urements, Livezey 1989), Royal and Rockhopper Penguins are, penguins: ecology and management. Chipping Norton: Sur- like all the crested penguins, among the most dimorphic, and rey Beatty & Sons. pp. 299–318. can be sexed reliably in the field using these characters. AMAT, J.A., VINUELA, J. & FERRER, M. 1993. Sexing Chinstrap Penguins (Pygoscelis antarctica) by morphologi- Inter-population comparisons cal measurements. Colonial Waterbirds 16: 213–215. BARRETT, R.T., PETERZ, M., FURNESS, R.W. & DURINCK, Comparisons of the Royal Penguin data presented in this study J. 1989. The variability of biometric measurements. Ring. and those in Woehler’s (1995) Macquarie Island study indi- Migr. 10: 13–16. cate significant differences in bill depth for males, and bill CARRICK, J. 1972. Population ecology of the Australian length for both sexes. In this study, male bill depth, male bill Black–backed Magpie, Royal Penguin, and Silver Gull. length and female bill length were 92.5%, 106.0% and 106.7% Wildl. Res. Rep. 2: 41–99. the size of individuals measured in Woehler’s (1995) study, DAVIS, L.S. & SPIERS, E.A.H. 1990. Mate choice in pen- respectively. Woehler’s (1995) study was carried out at a dif- guins. In: Davis, L.S. & Darby, J.T. (Eds). Penguin biology. ferent colony than this study, and whereas one cannot discount San Diego: Academic Press. pp. 377–397. the possibility that there are morphological differences be- GALES, R. 1988. Sexing adult Blue Penguins by external tween the various colonies on the island, the most likely cause measurements. Notornis 35: 71–75. is differences in measurements taken between various work- GROSCOLAS, R. 1990. Metabolic adaptations to fasting in ers (e.g. Barrett et al. 1989, Lorentsen & Rov 1994). Whereas Emperor and King Penguins. In: Davis, L.S. & Darby, J.T. this study and that of Woehler’s (1995) followed the techniques (Eds). Penguin biology. San Diego: Academic Press. pp. of Warham (1975), there were subtle differences in the inter- 269–296. 1996 Hull: Morphometric indices for sexing penguins 27

HULL, C.L. & WILSON, J. 1996. The effect of investigators Magellanic Penguins by discriminant analysis of morpho- on the breeding success of Royal, Eudyptes schlegeli and metric measurements. Colonial Waterbirds 10: 50–54. Rockhopper E. chrysocome, Penguins at Macquarie Island. SMITH. G.T. 1970. Studies of the behaviour and reproduction Polar Biol. 16: 335–337. of the Royal Penguin Eudyptes chrysolophus schlegeli. PhD KERRY, K.R. AGNEW, D.J., CLARKE, J.R. & ELSE, G.D. Thesis. Canberra: Australian National University. 1992. Use of morphometric parameters for the determina- TABACHNIK, B.G. & FIDELL, L.S. 1989. Using multivariate tion of sex of Adélie Penguins. Wildl. Res. 19: 657–664. statistics. New York: Harper Collins Publishers. LIVEZEY, B.C. 1989. Morphometric patterns in recent and WARHAM. J. 1963. The Rockhopper Penguin, Eudyptes fossil penguins (Aves, Sphenisciformes). J. Zool., Lond. chrysocome, at Macquarie Island. Auk 80: 229–256. 219: 269–307. WARHAM, J. 1971. Aspects of breeding behaviour in the LORENTSEN, S.-H. & RØV, R. 1994. Sex determination of Royal Penguin Eudyptes chrysolophus schlegeli. Notornis Antarctic Petrels Thalassoica antarctica by discriminant 18: 91–115. analysis of morphometric characters. Polar Biol. 14: 143– WARHAM, J. 1972. Breeding seasons and sexual dimorphism 145. in Rockhopper Penguins. Auk 89: 86–105. MARCHANT, S. & HIGGINS, P.J. 1990. Handbook of Aus- WARHAM, J. 1975. The crested penguins. In: Stonehouse, B. tralian, New Zealand and Antarctic Birds. Volume 1 Ratites (Ed.). The biology of penguins. London: MacMillan Press. to Ducks. Part A Ratites to Petrels. Melbourne: Oxford Uni- pp. 189–269. versity Press. WOEHLER, E.J. 1995. Bill morphology of Royal and Maca- MURIE, J.O., DAVIS, L.S. & McLEAN, I.G. 1991. Identify- roni Penguins, and geographic variation within eudyptid ing the sex of Fiordland Crested Penguins by morphometric penguins. In: Dann, P., Norman, I. & Reilly, P. (Eds). The characters. Notornis 38: 233–238. penguins: ecology and management. Chipping Norton: Sur- SCOLARO, J.A. 1987. Sexing fledglings and yearlings of rey Beatty & Sons. pp. 319–330.

Rockhopper Penguin by Liz McMahon (Harrison, J.A. et al. 1997. The atlas of southern African birds)